On Tue, Jul 23, 2019 at 10:02:56AM +0200, Jean Delvare wrote: > I think you are contacting the wrong people and list. The iTCO_wdt > driver is a watchdog driver and as such it is handled by the watchdog > subsystem maintainers (Wim Van Sebroeck and Guenter Roeck). You better > write to them and their list (linux-watchdog@vger). Hello and yes, I just came from there as we could not work it out :) https://www.spinics.net/lists/linux-watchdog/msg16232.html > The fact that iTCO_wdt loads when you load i2c-i801 is just a side > effect of an implementation detail (i2c-i801 instantiates the watchdog > device in certain cases, then module alias magic gets the needed driver > loaded automatically). If you only care about SMBus and not watchdog, > you can ignore the error message completely. > > If you care about the watchdog feature, you should check what is > conflicting with iTCO_wdt. Error -16 is -EBUSY which suggests that > another driver has already grabbed the memory range. This should be > visible in /proc/iomem. > > If it was grabbed by ACPI, you may need an ACPI driver for your > watchdog (if anything like that exists). If it was grabbed by another > non-ACPI driver, both drivers may have to be modified to synchronize > their access to the hardware. The only conflict we could find is that loading i2c-i801 creates this entry in /proc/iomem: 00c5fffc-00c5ffff : iTCO_wdt and loading iTCO_wdt after that produces the -16 error. Loading iTCO_wdt before i2c-i801 does not produce an error but also does not create /dev/watchdog. If i2c-i801 is loaded after that the same -16 error appears. I don't think any other watchdog module is running because /dev/watchdog does not exist, but I can send a lsmod output. So the main question at the moment is why loading just i2c-i801 produces an iTCO_wdt entry in /proc/iomem? Could it be that both i2c-i801 and iTCO_wdt are trying to register the same region? -- Valentin
